Making historical toxicology data work harder in safety assessment

Safety assessment is a fundamental part of drug development. Before a potential therapy reaches clinical trials, researchers need to understand its safety profile and identify potential toxicities. However, the demands placed on toxicology are changing.

Drug developers are expected to deliver programmes more quickly, improve the relevance of nonclinical findings to humans and reduce reliance on in vivo studies. Advances in artificial intelligence (AI), digital platforms and data science are also creating new opportunities to rethink how safety studies are designed and interpreted.

One development attracting attention is the use of Virtual Control Groups (VCGs), which use carefully curated historical control data to replace, reduce or augment concurrent control groups in selected toxicology studies.

Drug Target Review spoke with Laura Lotfi, Digital Scientific Innovation and Product Management Leader at Charles River Laboratories, about the challenges facing modern safety assessment, the growing role of advanced analytics and why making better use of existing data could change how toxicology is conducted.

Why safety assessment is under pressure

Safety assessment is facing several challenges simultaneously.

Drug developers want to accelerate timelines without compromising scientific quality, while regulators and the wider research community continue to encourage new approaches that reduce animal use and improve the translation of non-clinical findings into human outcomes.

One of the biggest obstacles is not necessarily generating more data but making better use of the data that already exists.

“The industry has generated high-quality toxicology data, yet much of it remains fragmented, inconsistently structured, or inaccessible for advanced analysis,” Lotfi says. “This limits our ability to contextualise findings or build predictive frameworks.”

The industry has generated high-quality toxicology data, yet much of it remains fragmented, inconsistently structured, or inaccessible for advanced analysis.

Variability also remains a challenge. Biological differences, operational variation and inconsistencies between studies can all make results more difficult to interpret and reproduce.

Using historical data to strengthen studies

Advances in digital platforms and data analytics are beginning to change how toxicology studies are interpreted.

Traditionally, researchers have evaluated findings against the concurrent control group included within a single study. Today, much larger curated targeted datasets can provide additional context, helping scientists understand whether an observed finding falls within expected biological variation or represents a genuine treatment-related effect.

“Advanced analytics is moving toxicology toward a more contextual and predictive paradigm,” Lotfi says.

“Importantly, this is not about replacing scientific judgment but augmenting it,” Lotfi explains. “The combination of domain expertise and data-driven insights leads to more robust conclusions and reduces uncertainty.”

Over time, studies could become more adaptive, with data informing decisions throughout a project rather than only after it has been completed.

What are Virtual Control Groups?

Virtual Control Groups (VCGs) use carefully curated historical control data from previous toxicology studies to replace, reduce or augment concurrent control groups under comparable study conditions.

VCGs are not simply a measure to save costs or resources – they enhance scientific context.

Developing a VCG involves far more than combining historical datasets. Studies must be carefully matched for factors such as but not limited to study design and conditions, animal strain, route of administration and endpoint domains before meaningful comparisons can be made. Historical data also requires rigorous curation, standardisation and validation.

“VCGs are not simply a measure to save costs or resources – they enhance scientific context,” Lotfi explains.

Their implementation also depends on robust digital infrastructure. Data integrity, traceability and governance are all essential if VCGs are to gain wider acceptance within regulated environments.

Supporting the 3Rs

For Lotfi, one of the most significant aspects of VCGs is their potential to support both scientific quality and the principles of the 3Rs – replacement, reduction and refinement of animal use in research.

“This is a pivotal development because it challenges a long-standing assumption that reducing animal use inherently compromises scientific rigour,” she says. “VCGs demonstrate that, with the right data foundation and validation, we can achieve both.”

Using larger, well-characterised historical datasets can strengthen the interpretation of safety findings while reducing the need for additional concurrent control animals. Better-informed safety decisions early in development could help researchers focus resources on the most promising candidates while avoiding unnecessary follow-up studies.

Building confidence through collaboration

Scientific evidence alone will not determine whether methodologies such as VCGs become routine practice.

“Scientific innovation alone is not sufficient – regulatory confidence and alignment are critical for adoption,” she says.

Regulators will need clear evidence that VCGs can be applied consistently and transparently across different study types. That means being able to demonstrate how historical data has been selected, how the methodology has been validated and where its limitations lie.

Scientific innovation alone is not sufficient – regulatory confidence and alignment are critical for adoption.

Building that evidence will require collaboration across the industry. No single organisation is likely to generate enough data to establish widespread confidence on its own, making shared data standards, common validation approaches and collective experience essential.

If those foundations can be established, VCGs are more likely to become a routine part of safety assessment rather than remaining a promising research methodology.

Digital platforms and AI

Although AI has become one of the dominant topics across drug discovery, Lotfi says its value depends on the digital platforms and data that support it.

Digital platforms provide the infrastructure needed to organise, integrate and analyse data, while AI helps researchers identify patterns that would be difficult to detect manually.

In safety assessment, these technologies can support feasibility assessments, improve study design and reduce unnecessary experimentation by making better use of existing information.

“AI is not a standalone solution – it is most powerful when combined with domain expertise and embedded within robust digital ecosystems,” Lotfi explains.

She believes these technologies can also contribute to more ethical research by maximising the value of data that has already been generated. Rather than collecting additional information unnecessarily, researchers can use existing datasets to support more targeted, hypothesis-driven studies.

Where safety assessment goes next

VCGs illustrate how historical toxicology data can become an active part of study design rather than simply providing a reference once studies have been completed. Combined with advanced analytics, they give researchers another way to interpret findings, improve study planning and reduce the need for concurrent control animals.

Whether these methods become part of routine safety assessment will depend on continued validation and evidence that they can be applied consistently across different study types. Regulators will need confidence not only in the quality of the historical data, but also in how it has been selected, standardised and analysed.

For organisations investing in digital safety assessment, this means thinking beyond individual studies. Historical toxicology data is only useful if it can be found, compared and reused. Laboratories that invest in curating and standardising those data today will be better positioned to apply methodologies such as VCGs in future studies.